Matplotlib多纵轴（寄生轴）散点图¶

发布于：2019-05-22 | 分类：python/vba/cpp

---

我们知道Matplotlib散点图的一个纵轴上可以画多条曲线，当纵轴数据属于不同类型（例如温度和压力）或者不同尺度，那就需要更多的纵轴来分别容纳。这类似于Excel图表中的次坐标轴，只不过Matplotlib可以呈现更多的纵轴。本文基于Matplotlib官网给出的实例封装了一个实现寄生轴散点图的函数。

概述¶

这些纵轴y axes共享同一横轴x axes数据，因此有两种思路可以实现：

• 孪生轴matplotlib.axes.Axes.twinx，基于同一个x轴但各自独立的一系列轴
• 寄生轴，包括一个主轴parasite_axes.HostAxes和从主轴构造的多个寄生轴parasite_axes.ParasiteAxes

Matplotlib官网给出了三个实例，分别采用了普通孪生轴，基于parasite_axes下host_subplot的孪生轴，以及parasite_axes下寄生轴的方法：

1. Multiple Yaxis With Spines
2. Demo Parasite Axes2
3. Parasite Axes demo

代码¶

本文基于parasite_axes寄生轴的方式封装了如下函数：

'''
Create a parasite axes, which would share the x scale with a host axes, 
but show a different scale in y direction.

This approach uses the parasite_axes' HostAxes and ParasiteAxes referenced from:
https://matplotlib.org/3.1.1/gallery/axisartist/demo_parasite_axes.html

params:
    xdata: 1d numpy array
    ydata: a list of structured array, each array represents one or more curves in associated y-axis.
        the field name of structured array become curve label. A reference to structed array:
        https://docs.scipy.org/doc/numpy/user/basics.rec.html
    xlabel: x label, `X` by default
    ylabels: list of label for each parasite axes, Y-0, Y-1 by default
    title: figure title
    colors: a list of colors for each curve, jet colormap by default
    markers: a list of markers for each curve, none by default
    linestyles: a list of line styles for each curve, solid line ('-') by default
    savefile: save figure if true else plot only

    colors, markers, linestyles are ignored if the length less than the count of all curves.
'''

import matplotlib.pyplot as plt
from mpl_toolkits.axisartist.parasite_axes import HostAxes, ParasiteAxes
from matplotlib import cm
import numpy as np


def parasite_axes(xdata, ydata, xlabel=None, ylabels=None, title=None, colors=None, markers=None, linestyles=None, savefile=None):

    fig = plt.figure(figsize=(10,7))    

    # host axis
    # - span for plotting multi-y axes
    # - hide axes
    span = 0.05*len(ydata) + 0.01
    host = HostAxes(fig, [span, 0.1, 0.88-span, 0.8]) # left, bottom, width, height
    host.axis['left','right'].set_visible(False)
    host.axis['top'].toggle(all=False)    
    host.set_xlabel(xlabel if xlabel else 'X')
    host.set_xlim(int(min(xdata)), int(max(xdata)))

    # create parasite axes
    para_axes = [ParasiteAxes(host, sharex=host) for _ in ydata]        

    # count of curves in all axes
    num = sum(len(data.dtype.names) for data in ydata)
    if not ylabels:
        ylabels = ['Y-{0}'.format(i) for i in range(len(ydata))]
    if not colors or len(colors)<num:
        colors = [cm.jet(x) for x in np.linspace(0.0, 1.0, num)]
    if not linestyles or len(linestyles)<num:
        linestyles = ['-'] * num
    if not markers or len(markers)<num:
        markers = [''] * num

    curve_index = 0
    for i, (par, ylabel, data) in enumerate(zip(para_axes, ylabels, ydata)):
        host.parasites.append(par)

        par.axis['new_left'] = par._grid_helper.new_fixed_axis(
            loc='left', 
            axes=par, 
            offset=(-35*i, 0))

        # add curve
        for curve_name in data.dtype.names:
            par.plot(xdata, data[curve_name], 
                label=curve_name, 
                color=colors[curve_index], 
                linestyle=linestyles[curve_index],
                marker=markers[curve_index],
                linewidth=0.5)
            curve_index += 1

        # label
        par.set_ylabel(ylabel)
        par.axis['left'].label.set_pad(0)

        # limits and ticks
        # numpy_array = data.view((data.dtype[0], len(data.dtype.names)))
        numpy_array = np.array(data.tolist())
        y_min, y_max = np.min(numpy_array), np.max(numpy_array)
        ylim_min = round(y_min-0.1*(y_max-y_min), 1)
        ylim_max = round(y_max+0.2*(y_max-y_min), 1)
        par.set_ylim(ylim_min, ylim_max)
        par.set_yticks(np.linspace(ylim_min, ylim_max, 5))
        par.axis['new_left'].major_ticklabels.set_axis_direction('top')

    # grid on
    para_axes[0].grid(True, linestyle='--')
    para_axes[0].axis['right'].set_visible(True)

    host.set_title(title if title else 'No Title')   
    host.legend(loc='upper left', bbox_to_anchor=(1.01, 1))
    fig.add_axes(host)

    if not savefile:
        plt.show()
    else:
        plt.savefig(savefile, dpi=120)



if __name__ == '__main__':

    xdata = np.linspace(1, 10, 20)
    ydata = [None] * 3
    ylabels = []
    colors, linestyles, markers = [], [], []

    # y axis 1
    ydata[0] = np.core.records.fromarrays([np.exp(xdata/10)], names='np.exp(x/10)')
    ylabels.append('exponential function')
    colors.append('r')
    linestyles.append('-')
    markers.append('')

    # y axis 2
    ydata[1] = np.core.records.fromarrays([np.sin(xdata), np.cos(xdata)], names='y=sin(x), y=cos(x)')
    ylabels.append('trigonometric functions')
    colors.extend(['g']*2)
    linestyles.extend(['-', '--'])
    markers.extend(['', 'o'])

    # y axis 3
    ydata[2] = np.core.records.fromarrays([xdata**1.5, xdata, xdata**0.5], names='y=x^1.5, y=x, y=x^0.5')
    ylabels.append('power functions')
    colors.extend(['b']*3)
    linestyles.extend(['-', '--', '-.'])
    markers.extend(['', 'o', '^'])    

    parasite_axes(xdata, ydata, title='Sample Parasite Axes', xlabel='sample x', 
        ylabels=ylabels, colors=colors, linestyles=linestyles, markers=markers)

测试实例显示结果如下：